Cancer-Associated Adipocytes Exhibit an Activated Phenotype and Contribute to Breast Cancer Invasion

Cells that compose the tumor stroma are associated, if not obligate, partners in tumor progression (1–3). In breast cancer, stromal cells include resident adipocytes and fibroblasts, a wide range of recruited hematopoietic cells, and newly formed blood vessels with their associated cells (3). Dynamic and reciprocal communication between epithelial and stromal compartments occurs during breast cancer progression (1, 2). Cancer cells usually go about generating a supportive microenvironment by activating the wound-healing response of the host (4). Conversely, the stromal cells, such as cancer-associated fibroblasts or tumor-associated macrophages, promote tumor progression by secreting growth factors, chemokines, and promigratory extracellular matrix components (2). To date, most of the studies focused on cancer cell–mesenchymal cell interactions have emphasized the contribution of fibroblasts, endothelial, and inflammatory cells (1). Little attention has been given to the mature adipocytes despite the fact that in breast cancers, early local tumor invasion results in immediate proximity of cancer cells to adipocytes (3). Until recently, adipocytes were mainly considered as an energy storage depot, but there are now clear evidences pointing adipocytes as endocrine cells producing hormones, growth factors, cytokines, and other molecules named adipokines (5). Therefore, mature adipocytes represent excellent candidates to influence tumor behavior through heterotypic signaling processes.

Adipokines, such as leptin, adiponectin (APN), hepatocyte growth factor, or collagen VI, stimulate the growth and survival of breast tumor cells even in estrogen-negative cells (6, 7). Moreover, it has been shown that metalloproteinase (MMP)-11/stromelysin-3 is induced in adipose tissue by cancer cells as they invade their surrounding environment (8). Obesity, wherein the normal balance of adipose tissue secretory proteins is disturbed, has been recently identified as a negative prognosis factor for breast cancer (9, 10). This poor prognosis seems to be independent of menopausal status, tumor stage, and tumor hormone–binding characteristics (for review, see refs. 11–13). Several studies pointed out that obese women exhibit at diagnosis an increase in lymph nodes involvement and a higher propensity to distant metastasis (10, 11, 14–17). Furthermore, obesity is also associated with the occurrence of tumors of high grade with increased local and distant invasion in prostate cancer (18). Taken together, these compelling observations suggest that excess adiposity may favor tumor invasiveness by yet uncharacterized mechanism(s). We show here that tumor-surrounding adipocytes exhibit profound phenotypic changes in vitro and in human tumors, and these modified adipocytes promote tumor cell invasion and metastasis.

The murine 3T3-F442A preadipocyte cell line was differentiated as described previously (19). To estimate the adipocyte phenotype, triglyceride (TG) content was quantified using a colorimetric kit (Wako Chemicals), and red oil staining was carried out as previously described (20). The human breast cancer cell line ZR 75.1 (provided by Prof. K. Bystricky, Laboratoire de Biologie Moléculaire Eucaryote, Toulouse, France) and the murine 67NR and 4T1 breast cancer cell lines (ref. 21; provided by Dr. S. Vagner, INSERM U563, Toulouse, France) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (ZR 75.1 and 67NR) and 5% (4T1) fetal calf serum (FCS). The human breast cancer cell line SUM159PT (provided by Dr. J Piette, IGMM, Montpellier, France) was grown in Ham F12 (50:50) medium complemented with 5% FCS, 1 μg/mL hydrocortisone (Sigma), and 0.2 UI/mL insulin. All the cell lines were used within 2 months after resuscitation of frozen aliquots. Cell lines were authenticated on the basis of viability, recovery, growth, and morphology. For ZR 75.1 and SUM159PT cell lines, the expression status of estrogen receptor (ER) α, E-cadherin, N-cadherin, and vimentin was confirmed by immunoblotting before they were used in the experiments (22). For all cell lines, their invasive behavior was confirmed using Boyden chamber assays before they were used in the experiments. Cell lines 67NR and 4T1 were also tested for their ability to form (4T1) or not (67NR) metastatic tumors when injected in syngenic BALB/c mice (21).

Breast adipose tissue samples were collected from reduction mammoplasty (RM) or tumorectomy (TU), according to the guidelines of the Ethical Committee of Toulouse-Rangueil. All subjects gave their informed consent to participate in the study, and investigations were conducted in accordance with the Declaration of Helsinki as revised in 2000. In cases of TU, the fat tissue samples obtained were immediately close to the tumor mass and were devoid of fibrosis, as assessed macroscopically (data not shown). Adipose tissue pieces were immediately used for collagenase digestion as previously described (23) and centrifuged (20 minutes, room temperature, 1,000 rpm) to separate adipocytes from the stroma vascular fraction (pellet, SVF).

Tumor cells and adipocytes were cocultured using a Transwell culture system (0.4-μm pore size; Millipore). A total of 5 × 10⁴ (67NR), 1.5 × 10⁵ (SUM159PT, 4T1), or 3 × 10⁵ (ZR 75.1) cells were seeded in the top chamber of the Transwell system in the culture medium of adipocytes and cocultivated with or without mature adipocytes in the bottom chamber for the indicated times. Adipocytes or tumor cells cultivated alone in similar conditions served as controls. To evaluate the effect of adipocyte-conditioned medium (Ad-CM) on tumor cell invasion, mature adipocytes were cultivated overnight with serum-free medium containing 1% of delipidated bovine serum albumin (Sigma) and medium was collected. CM was also collected from adipocytes previously cocultivated during 3 days with tumor cells [cancer-associated adipocytes conditioned medium (CAA-CM)]. After 3 days of coculture in the presence of mature adipocytes, Ad-CM, or CAA-CM (complemented with 10% FCS), tumor cells were used to conduct Matrigel invasion assays as previously described (24). A total of 35 to 40 fields per filter were counted using an optical microscope. Invasion assays were also conducted with tumor cells previously cultivated for 3 days with recombinant human IL-6 (PeproTech).

Whole cell extracts and immunoblots were prepared as previously described (25). All the antibodies used in our study are presented in Supplementary Information.

A total of 1 × 10⁶ 4T1 cells, previously cocultivated either with or without mature adipocytes for 4 days, were injected intravenously in BALB/c mice (6 animals per condition). Mice were sacrificed after 14 days, and lungs were taken for analyses. Nodules were scored from 1 to 5 according to their likelihood to represent metastasis to lung (0, normal lung tissue; 1, tissue with 1–5 nodules; 2, 6–10 nodules; 3, 11–20 nodules; 4, 21–30 nodules; and 5, >31 nodules). Lungs were then fixed in 4% paraformaldehyde and embedded in paraffin. Sections were stained with hematoxylin and eosin (H&E) and nodules were observed. The mean number of nodules per lung area and the mean nodule area as a function of lung area (that defined the nodule area index) were measured by ImageJ software program. The local Institutional Animal Care and Use Committee approved the experimental protocols described in the study.

Immunofluorescence and immunohistochemistry were carried out as previously described (8, 24). Concentration of secreted factors in supernatants was determined using the xMAP Technology (Luminex), with a Milliplex kit (Millipore).

Gene expression was analyzed using real-time PCR as described previously (23). The sequences of the primers are presented in Supplementary Information.

Quantitative expression of biological markers measured by quantitative PCR (qPCR) in isolated human adipocytes was assessed using estimations of means and SDs. Hence, a normal distribution of these measures was assumed. However, the size of subpopulations was limited in most cases. Thus, we tested differences about the expression of biological markers between subpopulations of interest, using nonparametric tests (i.e., Wilcoxon's nonparametric rank test). Quartiles were also computed; box plots permitted to appreciate the presence of outliers and to illustrate differences in the expression of biological markers. The threshold of statistical significance was fixed at alpha risk at 5%. SPlus statistical software was used for the data analysis (26). For the in vitro experiments, the statistical significance of differences between the means (at least 3 independent experiments) was evaluated using Student's t test, done using Prism (GraphPad Inc.). The values of *P ≥ 0.05, **P < 0.01, and ***P < 0.001 were deemed as significant whereas “NS” stands for not significant.

Human cancer cells with low (ZR 75.1) or high (SUM159PT) invasive capacities (27) were grown for 3 days on Transwells, allowing the diffusion of soluble factors, in the presence or absence of murine mature adipocytes (28). After 3 days, tumor cells were trypsinized and Matrigel invasion assays were conducted in a medium containing either 0% or 10% FCS. As shown in Figure 1A, the invasive capacity of ZR 75.1 tumor cells in 10% FCS containing medium was increased by 3-fold in cells previously cocultivated for 3 days with adipocytes as compared with tumor cells grown alone (P < 0.01). No additional increase in the invasive capacity of ZR 75.1 was observed when the time of coculture with adipocytes was increased up to 5 days (data not shown). This effect was not observed when tumor cells were cocultivated with preadipocytes (Supplementary Fig. S1A). Similar results were obtained with the SUM159PT cell line, although to a lesser extent (Fig. 1A, P < 0.05). Experiments were also carried out with the mouse mammary cell line 4T1 and its nonmetastatic sibling cell line 67NR (21). Coculture with adipocytes increases both 67NR and 4T1 invasive capacities by 2-fold (Fig. 1B, P < 0.05 and 0.01, respectively). Therefore, mature adipocytes are able to stimulate the invasive capacities of murine and human breast cancer cell lines that are either positive (ZR 75.1) or negative (SUM159PT, 67NR, 4T1) for the ER. In addition, the proinvasive effect was independent from the tumor growth–promoting effect because coculture with adipocytes has no effect on ZR 75.1 or murine tumor cell lines proliferation in contrast to SUM159PT cells (Supplementary Fig. S1B). When tumor cells were grown in the presence of CM obtained from adipocytes, which have never been cocultivated with tumor cells (Ad-CM), invasive capacities were not significantly increased for any of cell lines (Fig. 1C). When CM was prepared from adipocytes previously cocultivated with tumor cells (CAA-CM), a significant stimulation of tumor invasive capacities was observed in all the tested models as compared with cells grown in control medium (P < 0.05 for the 67NR cells, P < 0.01 for the ZR 75.1, SUM159PT, and 4T1 cells). These results highlight that a cross-talk between the 2 cell populations is necessary to obtain this effect (Fig. 1C).

Noncocultivated ZR 75.1 cells displayed a characteristic epithelial morphology and form compacted colonies. Yet, when these cells were cocultivated with adipocytes from 3 to 5 days, morphologic changes became apparent. Cells formed less compacted colonies and exhibited a more scattered aspect. In addition, these cells displayed a downregulation of membrane E-cadherin expression, in association with β-catenin redistribution, resulting into a disorganized state in the cytoplasm (Fig. 2A). Decreased E-cadherin expression was observed both at mRNA (Fig. 2B) and protein levels (Fig. 2C) without an increased expression of mesenchymal markers (Fig. 2B and C), suggesting the occurrence of an incomplete epithelial-mesenchymal transition (EMT). For the SUM159PT cell line, which has already undergone a complete EMT, morphologic changes were also observed along with the reorganization and polarization of vimentin (Fig. 2A) without a change in its expression level (Fig. 2C).

Our coculture system shows that adipocytes stimulate the invasiveness of tumor cells in vitro. To test the hypothesis that “priming” tumor cells with adipocytes regulate tumor cell distal organ seeding and growth, we conducted tail vein metastasis assays, using the 4T1 cell line. 4T1 tumor cells were cultured in vitro for 4 days in the presence or absence of adipocytes, harvested, and injected intravenously into BALB/c mice. As shown in Figure 3A, the number of lung metastases was enhanced in mice injected with 4T1 cells previously cocultivated with adipocytes as compared with mice injected with 4T1 grown alone (P < 0.05). Histologic analysis (Fig. 3B) confirmed that both the number and size of the nodules were significantly increased in mice injected with adipocytes-cocultivated 4T1 compared with 4T1 cells grown alone (Fig. 3C and D, P < 0.05 for the 2 parameters). Therefore, our results clearly show that adipocytes promote the invasive phenotype of breast tumor cells both in vitro and in vivo.

Adipocytes have been shown to promote tumor survival and growth in vitro and in vivo (6, 29), but their contribution to tumor cell invasion is an emerging concept. Because our results showed that a cross-talk between the 2 cell populations is necessary to observe the proinvasive effect (Fig. 1C), we next examined the phenotype of cocultivated adipocytes. As shown in Figure 4A (left), mature adipocytes cocultivated with ZR 75.1 tumor cells exhibited a dramatic reduction in the number and size of lipid droplets that correlated with a significant decrease in lipid accumulation (Fig. 4A, right, P < 0.05). These adipocytes lose several terminal differentiation markers exemplified by the strong decrease in HSL (hormone-sensitive lipase), APN, resistin, and aP2 mRNA (ref. 5; Fig. 4B, P < 0.05 for all the markers), this loss was also confirmed at protein levels for APN and resistin (Fig. 4C, P < 0.05). In cocultivated adipocytes, the level of C/EBPα was dramatically reduced (P < 0.001) whereas that of PPARγ was not significantly decreased (Fig. 4B). As previously stated, MMP-11/stromelysin-3 is induced in vivo in adipose tissue by cancer cells as they invade their surrounding environment (8). Similarly, an upregulation for this protease was noted in the cocultivated adipocytes, highlighting the in vivo relevance of this coculture model (Fig. 4B, P < 0.001). In contrast, the levels of MMP-9 (Fig. 4B) and MMP-2 (not shown) remain unchanged whereas a significant increase (10-fold) of plasminogen activator inhibitor-1 (PAI-1) expression was found (Fig. 4B, P < 0.01). In cocultivated adipocytes, a 5-fold increase in levels of IL-6, IL-1β, and TNF-α was shown at mRNA levels (Fig. 4B, P < 0.001 for all cytokines). At protein levels, however, only IL-6 and IL-1β were found to be significantly higher in the supernatant of cocultivated cells (Fig. 4C, P < 0.05 for both cytokines) whereas the levels of TNF-α remained undetectable in both conditions (data not shown). Of note, these extensive adipocyte phenotypic changes were obtained with other tumor cells lines (Supplementary Fig. S2). Furthermore, these results were confirmed with mammary adipocytes derived from ex vivo differentiation of progenitors purified from mammary adipose tissue of healthy donors undergoing RM (6 independent donors; Supplementary Fig. S3). Therefore, we extensively described tumor-induced changes in adipocytes that are reproducibly displayed by both human and mouse cocultivated adipocytes. Collectively, our observations clearly show that cocultivated adipocytes generally exhibit a loss of lipid content, a decrease in late adipose markers, and overexpression of inflammatory cytokines and proteases. We named these cells “cancer associated adipocytes” (11).

We next investigate whether these cells are present in primary human tumors. To address this critical issue, we first conducted a histologic analysis of human breast tumors. As exemplified in Figure 5A, we consistently observed at the invasive front of human breast tumors the presence of adipocytes of smaller sizes with a dilated interstitial space, which might be related to extensive extracellular matrix occurring in tumor-surrounding adipocytes such as collagen VI overexpression (6). Further characterization of these CAAs was done by immunohistochemistry. As shown in Figure 5B (i and ii), we observed a decreased expression of the adipocyte marker APN in CAAs as compared with normal mammary adipose tissue. Although MMP-11 and IL-6 were not detected in mature adipocytes of RM (Fig. 5B, iii and v, respectively), these proteins were expressed within the rim of cytoplasm of the adipocytes located proximally to the invasive cancer cells (Fig. 5B, iv and vi, respectively). To further assess the expression of these markers by a quantitative approach, adipocytes were isolated from fat samples associated with TU or RM (23). A series of 28 samples, with 14 samples in each group (Supplementary Tables S1 and S2 for the clinical characteristics of the patients), was used comparable with respect to age (mean: 51.8 ± 7.3 vs. 60 ± 13) and BMI (26.2 ± 2.8 vs. 25.2 ± 2.3). Adipocytes isolated from tumor samples overexpressed MMP-11 (P = 0.001), PAI-1 (P < 0.01), and IL-6 (P < 0.001), as compared with adipocytes purified from normal mammary gland, whereas the mRNA levels of IL-1β and TNF-α were not statistically different (Fig. 5C). The adipose tissue also contains stromal cells (present in the SVF), such as fibroblasts, endothelial cells, and adipocyte precursors. The levels of expression of MMP-11, PAI-1, IL-6, IL-1β, and TNF-α remained unchanged between normal and tumor-surrounding adipose SVF, highlighting that the observed increase in proinflammatory molecules and proteases is an adipocyte-dedicated event (Fig. 5C). Furthermore, when the expression levels of IL-6 in adipocytes were analyzed as a function of tumor characteristics, we found that higher levels of this cytokine were associated with the tumors of larger size and with lymph nodes involvement (Fig. 5D). This correlation was not found either for PAI-1 or for MMP-11 (Supplementary Fig. S4). Taken together, these results show that CAAs are found in human tumors.

Our data strongly support the concept that an intimate cross-talk is established between cancer cells and mature adipocytes at the invasive front of the tumor. Invading tumor cells are able to modify adipocytes phenotype, which, in turn, stimulate cancer cells aggressive behavior (30). The mechanism(s) responsible for the transition from mature adipocytes to CAAs remain(s) unclear. Modification resembling to CAAs is also observed in adipose tissue of tumor-bearing mice far from the tumor site. In cachexia, adipocytes also exhibit morphologic modification, decreased in adipogenic transcription factors such as C/EBPα and late differentiation markers such as leptin (31). However, during cancer cachexia, adipocytes exhibit a decrease in resistin expression and upregulation of lipolytic enzymes such as HSL or ATGL (adipose triglyceride lipase) (31), events not found in CAAs (Fig. 4 and Supplementary Fig. S5), alluding to their unique phentotype. Several molecules expressed by tumor cells have been described as strong suppressors of adipogenesis such as IL-1β, TNF-α, or even IL-6 (32), representing one mechanism by which tumor cells might directly influence adipocyte behavior. An alternative mechanism might be adipocytes themselves as the source of autocrine signals that induce the CAAs phenotype. In favor of the latter hypothesis, we showed that tumor CM was not able to reproduce CAAs phenotype, suggesting that either adipocyte factors or inducible factors in tumor cells are necessary to observe this effect (Supplementary Fig. S6).

Our results show that IL-6 is overexpressed in tumor-surrounding adipocytes both in vitro and in vivo. We showed that IL-6 (at 10 ng/mL, a dose equivalent to the levels secreted by CAAs, Fig. 4C) significantly increased the invasive capacities in ZR 75.1 and 67NR cells (Fig. 6A, P < 0.01 for the 2 cell lines). As compared with untreated cells, the epithelial-like ZR 75.1 cells exposed to IL-6 form less compacted colonies and exhibit a more scattered aspect (Fig. 6B). A downregulation of membrane E-cadherin expression resulting into a disorganized state in the cytoplasm, in association with β-catenin redistribution (Fig. 6B), was observed in treated cells in accordance with the literature (33–35). Strikingly, when ZR 75.1 cells were cocultivated with adipocytes in the presence of murine IL-6 blocking antibody, the adipocyte proinvasive effect was significantly decreased (Fig. 6C, P < 0.01). mIL-6 blocking antibody specifically inhibits the proinvasive effect of CAA-CM in both 67NR and 4T1 models and had no effect on the invasive capacities of cells grown in control medium (Fig. 6D, P < 0.01 and P < 0.05, respectively). Altogether, these results indicate that IL-6 plays a key role in the CAAs proinvasive effects.

In breast cancer patients, the extent of the increase in serum IL-6 correlates with poor disease outcome and reduced prognosis (36). Recent evidence shows that fibroblast-derived IL-6 supports breast tumor growth and metastasis (37). Our results show that the stromal IL-6 is also produced by mature adipocytes surrounding the tumors. The absence of IL-6 overexpression in the SVF of human adipose tissue isolated from breast tumor biopsies (Fig. 5C) apparently contradicts a recent study conducted ex vivo, and not in vivo, using human adipose tissue–derived stromal cells cocultivated with tumor cells (38). Our results obtained in human tumors suggest that within the cellular complexity of the adipose tissue, the cross-talk with tumor cells is rather established with mature, than with precursors, adipose cells (Fig. 5C and D). In addition, because mature adipocytes are the most abundant cells in the breast tumor stroma (3), their cumulative secretion of IL-6 could therefore have profound impact on tumor progression.

Finally, this work paves the way for future clinical and cellular studies aimed at determining the impact of this cross-talk in obesity conditions. Indeed, our results suggest that the cross-talk between the CAAs and tumor components might be amplified in obesity in which the normal balance of adipose tissue secretory proteins is disturbed. Obesity is related to a condition of low chronic inflammation characterized by the abnormal production of inflammatory cytokines involved in insulin resistance (5). Obesity and the now-defined profile of CAAs share inflammation as common traits, including overexpression of IL-6 with the noticeable absence of overexpression of TNF-α in CAAs (Fig. 5). Interestingly, the association of high levels of IL-6 in CAAs with high tumor size and enhanced local invasion reflects those traits present in obese patients (10, 16, 17). Therefore, CAAs within a context of obesity should be more prone to amplify the negative cross-talk with tumor cells that we have identified and characterized in this study. This is a hypothesis that remains to be shown in both murine and human models. Such studies will provide unique opportunities to set up specific strategies for the treatment of the subsets of patients exhibiting aggressive diseases.

No potential conflicts of interest were disclosed.

The authors thank Sandrine Imbert for her excellent technical assistance, Sandra Grès for qPCR primer design, the Anexplo-Phenotypage platform, IFR150, in Toulouse for Luminex analysis, and all our colleagues from I2MR and IPBS for critical reading of the manuscript.

The “Canceropole Grand Sud-Ouest”, INCA (P. Valet, C. Muller, and G. Escourrou), the “Ligue Nationale Contre le Cancer (Comité “Midi-Pyrénées, Gers et du Lot” to C. Muller), and the radioprotection Committee of EDF (C. Muller) supported this work.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.